Benchmark vectorization

benchmark-vectorization.R

library(microbenchmark)
options(digits = 3)

# Initialize variables. 
vec <- 1:1e7  # initialize input vector, change size here
range <- seq_along(vec)  # for bare for-loop

sum_unallocated <- function(x){
  result <- c()
  for (i in seq_along(x)){
    result[i] <- x[i] + 1
  }
  return(result)
}

sum_allocated <- function(x){
  result <- vector(mode = "numeric", length=length(x))
  for (i in seq_along(x)){
    result[i] <- x[i] + 1
  }
  return(result)
}

sum_apply <- function(x){
  result <- sapply(x, FUN = function(y) { y + 1 })
  return(result)
}

sum_vectorized <- function(x){
  result <- x + 1
  return(result)
}

empty_for_loop <- function(x){
  for (i in seq_along(x)){ }  
}

bare_for_loop <- function(range){
  for (i in range){ }  
}

# Benchmark all results. 
microbenchmark(a <- sum_unallocated(vec), 
               b <- sum_allocated(vec),
               c <- sum_apply(vec),
               d <- sum_vectorized(vec), 
               empty_for_loop(vec),
               absolutely_bare_for_loop(range),
               times = 10L, unit = "s")
# Unit: seconds
# expr                          min      lq    mean  median      uq    max neval  cld
# a <- sum_unallocated(vec)  4.2620  4.3079  4.6143  4.5841  4.7909  5.317    10  c 
# b <- sum_allocated(vec)    0.6237  0.6508  0.7549  0.7036  0.7237  1.305    10  b  
# c <- sum_apply(vec)        9.3891 10.4658 11.1595 10.9814 12.0064 12.718    10  d
# d <- sum_vectorized(vec)   0.0261  0.0298  0.0974  0.0903  0.1554  0.214    10  a   
# empty_for_loop(vec)        0.0938  0.0952  0.0987  0.0995  0.1011  0.104    10  a   
# bare_for_loop(range)       0.0878  0.0883  0.0991  0.0918  0.0953  0.164    10  a  

# Check that all results are equal. 
all.equal(a, b)
all.equal(b, c)
all.equal(c, d)